Filter no more: A modified plankton sampler for rapid in‐water eDNA capture

The combination of an efficient sampling method and high‐throughput analysis of environmental DNA (eDNA) can be a powerful approach for characterising biodiversity across aquatic ecosystems. Plankton net tows are one of the oldest, simplest, and least expensive methods for seston and eDNA collection, but require laborious filtration steps which often lead to clogging and/or the introduction of contaminants. In this study, we used a cruising speed net (CSN) device enabling the collection of seston‐derived eDNA at 5 knots speed combined with a novel modified cod‐end with 20 μm nylon mesh inserts enabling eDNA capture while towing. We compared the performance of the CSN sampling protocol with the original conventional filtration of water sample versus the modified cod‐end. Samples were collected in parallel horizontal tows along New Zealand's North‐Eastern coastline. Concentrated water was filtered on conventional 5 μm cellulose acetate membranes, while the 20 μm nylon mesh inserts were immediately isolated post‐towing. Metabarcoding of bacterial 16S rRNA, eukaryotic nuclear 18S rRNA and mitochondrial COI genes, revealed no significant difference in alpha diversity between filtration techniques. In terms of community composition, a clear and significant shift could be observed between sampling sites and environments. Significant differences could be detected between filtration methods for 16S and COI markers, likely driven by fine‐scale differences at more turbid sheltered sites. Nonetheless, each technique could detect shifts in communities between sites and environments with similar sensitivity. Our results demonstrate the promising potential of the modified cod‐end to enable practical and cost‐effective isolation of eDNA‐derived biodiversity data from any vessel types (at ≤5 knots) across a large range of aquatic ecosystems and biogeographic scales.


| INTRODUC TI ON
The rapid collection and high-throughput molecular analysis of environmental DNA (eDNA) from aquatic ecosystems is a powerful new approach for in-depth biodiversity assessments (Deiner et al., 2017;Huerlimann et al., 2020;Ruppert et al., 2019).This is particularly important in a period of climate change affecting ecological stability at unprecedented rates across biomes (Cavicchioli et al., 2019;Hoegh-Guldberg & Bruno, 2010;Pörtner et al., 2022).
Given the immensity and highly dynamic nature of the world's aquatic systems and the tiny size and high turnover rate of eD-NA-derived planktonic communities, this is a great challenge that can only be addressed by more effective sampling across much larger spatio-temporal scales than are currently possible (Lauro et al., 2014).
Plankton nets are one of the oldest, simplest and least expensive methods of sampling seston (Gutkowska et al., 2012) and can considerably increase sampling volume and spatio-temporal coverage of eDNA material (Bucklin et al., 2021;Schabacker et al., 2020), but have two major limitations.First, they create considerable drag in the water column and cannot be towed at more than 2 knots, precluding their use on a range of vessel types or weather conditions.Von Ammon et al. ( 2020) developed and validated the use of a lightweight, easily deployable cruising speed net (CSN) that enabled the collection of ocean surface particulate material, micro-and mesoplankton while towed behind a boat at speeds up to 5 knots.The CSN consists of a specifically shaped PVC head, with a 20 cm wide isometric front eye, followed by a slightly conical 2 m long nylon net (20 μm pore size) ending with a 500 mL closed cod-end.This device is shaped to hydrodynamically sink and maintain its stability at 2-3 m depth sub-surface with minimal weights and lines to reduce drag and to facilitate ease of handling during sampling.The CSN effectively concentrates high amounts of seston-derived eDNA from the hundreds of litres of water passing through the device in a short timeframe, and can be used from a variety of vessel types, aquatic ecosystems or situations (e.g.citizen science programmes).Second, the concentrated sample accumulates in the cod-end during towing and must then be manually filtered on small (0.1-5 μm) membranes, a time-consuming process that often results in clogging and/or the introduction of contaminants (Andruszkiewicz et al., 2017;Goldberg et al., 2013;Turner et al., 2014).Although most eDNA studies consider that smaller pore sizes (<1 μm) maximise nucleic acids capture from a wide range of genetic material or environments (Capo et al., 2020;Collins et al., 2018;Kelly et al., 2014), recent evidence indicates that larger pore filters (5-20 μm) can be equally effective (Cooper et al., 2022;Egan et al., 2013;Sepulveda et al., 2019;Thomas et al., 2018;Turner et al., 2014;Wittwer et al., 2018).For example, Zaiko et al. (2022) showed no significant differences between membrane types (0.45, 1.2, and 5 μm) for capturing various fractions of target eDNA and eRNA (intact cells, broken cells, naked nucleic acids) from a model algal species, and demonstrated that larger pore size (5 μm) cellulose membranes offered the highest time-volume efficiency.
In this study, we engineered and tested a modified cod-end with 20 μm nylon mesh inserts for instantly concentrating the samples while towing, thus circumventing the active filtration step.We compared the performance of the CSN sampling protocol with conventional filtration (CF) versus in-water capture (IC), and validated the device through systematic metabarcoding analysis of bacterial (16S rRNA) and eukaryotic (18S rRNA and COI) markers.
We elaborate upon this new device and its ability to easily and more rapidly isolate eDNA-derived biodiversity data from aquatic

| In-water capture (IC) device
The modified cod-end or IC device was designed and manufactured by Sequench Ltd, Nelson, New Zealand.Made out of Trovidur© PVC plastic, the tubiform IC device is 93 × 145 mm in size and consists of two parts (Figure 1a).The inside of the larger front part is conical shape, starting with a straight in-flow opening of 77 mm in diameter (75 mm in length), and ending with a frustrum-shape out-flow with opening of 38 mm in diameter (Figure S1).The bottom part of the modified cod-end consists of a 93 × 23 mm screw cap that holds a 47 mm diameter eDNA nylon mesh insert.The filter holder is made of a perforated plastic grid that enables the in-water to pass through the nylon filter and escape the cod-end.The nylon mesh insert is locked in place by screwing the end cap to the main part of the modified cod-end (Figure 1a).

| Field sampling
We collected (14-15 June 2021) five sub-surface plankton samples using two different sampling methods at four sites corresponding to 20 stations in the East Auckland region, New Zealand, ranging from Waitematā Harbour to Hauraki Gulf (Figure 1b; Table S1) on-board the 22-meter-long SV Manawanui sailing vessel (Far Out Ocean Research Collective).
At each station, two Cruising Speed Nets (CSN) were deployed simultaneously from each side of the vessel and towed for exactly 3 min at 5 knots.One CSN was equipped with a 'conventional' 500 mL cod-end, and the other CSN with the 'modified' cod-end with 20 μm nylon mesh inserts able to capture the plankton concentrate while towing (Figure 2; Figure S1).High-grade 20 μm nylon mesh (NITEX™, Sefar Ltd., Switzerland) was cut in 50 × 50 mm pieces, UV-treated for 30 min and placed in sterilised bags until field work.
Following each tow, concentrated plankton samples from the 'conventional' cod-end were immediately filtered through 5 μm pore size cellulose acetate membranes (47 mm diameter, Sterlitech, WA, USA), following recommendations of Zaiko et al. (2022).Filters from both 'conventional' and 'modified' cod-ends were transferred into 2 mL vials containing RNA-Shield (Zymo, CA, USA) isolation buffer.All samples were stored at 4°C for 10 days, then at −20°C until further F I G U R E 2 Experimental design from eDNA collection to validation.Complete specifications and assembly of the cruising speed net as well as detailed methodology of the conventional filtration is presented in von Ammon et al. ( 2020).Image created using Biore nder.com.
analyses.Between sampling stations, all gears (CSN devices, codends, tweezers, filtering manifold cups) were soaked in 2% bleach solution for at least 10 min, then thoroughly rinsed with seawater from the sampling site to avoid cross contamination from other stations.Sterile gloves were worn throughout sampling and gear operations.For each sampling day, filter (5 and 20 μm) blanks were created by filtering (on-board the ship) 500 mL of ultrapure water in the manifold system and modified cod-end, respectively.
After demultiplexing fastq files and removing primers with cutadapt (version 4.2; Martin, 2011), sequences were quality filtered, denoised, merged and filtered from putative chimeric sequences using the default parameters of the 'DADA2' R package (version 1.21; Callahan et al., 2016).Amplicon sequence variants (ASVs) were taxonomically assigned using databases such as SILVA (16S and 18S rRNA genes), as well as MIDORI and GenBank (COI gene).Blanks were treated for potential contamination using the 'microDecon' R package (version 1.0.2;McKnight et al., 2019) and following rarefaction (Figure S2) samples with less than 1000 reads as well as rare ASVs (less than 1 reads in at least 2 samples, and with less than 10 reads overall) were removed from downstream analysis.
The taxonomic community composition was visualised with doughnut charts and richness per sample type was visualised using boxplots.The effect of sample type on richness was tested with one-way analysis of variance (ANOVA), followed by pairwise t-tests and beta-diversity was examined with non-metric multidimensional scaling (NMDS) plots and permutational analysis of variance (PERMANOVA) to test for differences between sites and environments.
Detailed methodologies on DNA extractions, PCR amplifications, sequencing, bioinformatics and biostatistics can be found in the Supplementary Material 1.The Cawthron Institute holds a Special Permit with the New Zealand Ministry for Primary Industries (SP822-2) that allows the taking of fish, aquatic life and seaweed for the purposes of education and investigative research.

| Sequence data and high-level biodiversity overview
The 40 biological samples (Table S1) were sequenced along with all controls and, following filtering (i.e., denoising, merging, and chimera, NUMT and putative contamination and rare variants removal), yielded a total of 167,432 sequence total reads or 4185 ± 320 reads per sample (553 ASVs) for the 16S rRNA, 816,284 sequence total reads or 20,407 ± 1566 reads per sample (204 ASVs) for the 18S rRNA and 554,647 sequence total reads or 13,866 ± 1197 reads per sample (234 ASVs) for the COI gene (Table S2; Figure S3).Doughnut charts revealed similar proportions of the top 10 bacterial and eukaryotic taxa (family level) between CF and IC sample types (Figure 3).

| Alpha and Beta diversity
Diversity analysis at lowest taxonomic (ASV) level showed that the mean observed richness was comparable between sample types (Figure 4a), with the one-way ANOVA revealing no significant difference across all markers (Table S3). Figure S3 shows that the majority of ASVs are shared between sample types (16S = 85.2%, 18S = 83.3%, and COI = 93.2%).
Beta diversity analyses revealed clear and linear community composition shifts from site 1 to site 4, a pattern held true across all marker genes, with noticeable clustering of stations from more sheltered/turbid environments in the Waitematā region (sites 1-2) versus open ocean stations in the Hauraki Gulf (sites 3-4) (Figure 4b-d).

| DISCUSS ION
Optimising eDNA sample collection is particularly relevant when targeting vast and logistically difficult to access aquatic ecosystems (Bowers et al., 2021).Overall, our results demonstrated the ability of the in-water capture (IC) method to rapidly isolate comparable prokaryotic and eukaryotic biodiversity consortia and richness from a range of surface seawater samples.Coupled with a CSN device, the new IC cod-end allows for the efficient accumulation and retention of eDNA material on a large (20 μm) filter pore size from approximately 1500 L of seawater within 3 min of vessel towing at 5 knots.
Conventional filtration is notoriously problematic when applied in the field or on-board of a vessel, as it requires immediate and substantial time effort in unsuitable conditions that may lead to sample contamination (Minamoto et al., 2016;Thomas et al., 2018;von Ammon et al., 2022).In this study, substantial contaminating sequences were recovered from field blanks despite the precautionary measures taken in our experimental design to minimise these.In particular, the 20 μm nylon field blanks yielded more contamination than the 5 μm cellulose acetate field blanks (Table S2), possibly due to insufficient UV-based treatment of the nylon mesh in this pilot study.This can be effectively remediated in future studies by using laboratory grade sterilised nylon membranes (Zaiko et al., 2022).
Despite operating in a state-of-the-art molecular laboratory, additional contamination was observed throughout the laboratory workflow for the bacterial 16S dataset which is difficult to avoid and a widely acknowledged occurrence in metabarcoding-based microbiome studies (Salter et al., 2014;Weiss et al., 2014).This highlights the fact that field-based eDNA studies should systematically include field and laboratory blanks at every step of operation and apply adequate post-sequencing contamination removal algorithms, as performed in the present study (Table S2).
Another challenge with the filtration process of aqueous eDNA is the difficult trade-off choosing between optimal filter pore size and maximum water volume that can most effectively capture nucleic acids from a range of environmental materials, from organisms' tissues to sub-cellular particles (Bowers et al., 2021;Deiner et al., 2015;Turner et al., 2014).While finer pores (0.1-3 μm) are more likely to retain microorganisms or free-floating nucleic particles, small water volumes may clog rapidly in turbid environments, often precluding comprehensive biodiversity assessments at local to regional scales (Andruszkiewicz et al., 2017;Goldberg et al., 2013).
The simple solution presented in this study, that is combining high water volumes afforded by the CSN, and accumulation and retention of fine particles onto the large (20 μm) nylon membrane of the IC cod-end, provides a great opportunity to bypass the cumbersome filtration step without losing in-depth biodiversity information across both bacterial and eukaryotic domains.
Overall, alpha and beta diversity analyses showed that CF and IC approaches yielded analogous biodiversity estimates, with congruent proportions of prokaryotic and eukaryotic taxa at family level, and no significant difference in richness at lowest (ASV) taxonomic level across all three molecular markers.In addition, the large majority of ASVs (>83%) were shared between sample types (Figure S3), confirming that IC captures similar biological assemblages from sub-surface seawater habitats than the CF method.Nevertheless, some fine-scale differences in community composition between sampling methods were observed for two markers (16S and COI), especially at the more turbid stations (sites 1 and 2) of the Waitematā navigation channel.One possible explanation is that the parallel tows of CSN devices from each side of the vessel captured slightly different assemblages due to higher micro-patchiness in biodiverse eutrophic habitats (Lunt & Smee, 2020;Priyadarshi et al., 2019) and the relatively high variability of the bacterial 16S and eukaryotic COI markers compared to the 18S gene (Pearman et al., 2020(Pearman et al., , 2021)).Another explanation could be that the visibly higher accumulation of seston material on the IC filters at these turbid locations (Figure S4) may have led to an increase of DNA molecules associated with sediment or mineral particles resulting in slightly different communities captured by these variable markers (Beng & Corlett, 2020).Finally, we cannot exclude the possibility that differences in filter pore size between methods had some influence on the capture efficiency of bacterial communities, although this was not the case at open ocean stations.Overall, we detected different communities at turbid sites but both methods were comparable in their ability to detect the effect of site and environment.In addition, CF and IC methods performed the same at sampling sites 3 and 4, indicating that current towing speed, duration and 20 μm nylon mesh size represent optimal settings for eDNA-based biodiversity assessments in coastal to open ocean environments.
In conclusion, we have developed and validated the use of a modified cod-end with nylon mesh inserts facilitating rapid in-water eDNA capture from large water volumes, enabling in-depth biodiversity screening from larger aquatic areas than currently possible while circumventing notoriously challenging conventional filtration processes.This new method contributes to a rapidly growing eDNA research field and the increasing need for fit-for-purpose nucleic acids sampling tools for global biodiversity assessments (Ruppert et al., 2019).A key asset of the IC device is its ability to minimise sample manipulation in remote areas, which is particularly relevant when sample collection is performed by non-expert end-users or citizens (von Ammon et al., 2020(von Ammon et al., , 2022)).Moreover, reducing the sample handling time before preservation from up to hours with conventional filtration to <3 min with IC reduces the risk of nucleic acid degradation, which is critical when dealing with environmental RNA (eRNA; Cristescu, 2019;Pochon et al., 2017).However, this pilot study also highlighted some limitations, such as fine-scale differences in community composition between sampling methods in turbid environments and the introduction of contaminants which require appropriate controls in field-based settings.Future work is required to further investigate eDNA and eRNA retention potential of the IC device under different velocity pressures and nylon mesh sizes from a range of aquatic environments.Additionally, the development of a miniaturized torpedo-like eDNA sampler, inspired by the 'Plankton Indicator' (Glover, 1953) but using the same in-water eDNA capture principle described here, could offer a less bulky and more versatile solution for large scale biodiversity, biomonitoring and citizen science applications in marine and freshwater ecosystems.
ecosystems.F I G U R E 1 (a) Manufacturing specifications and assembly of the in-water capture (IC) modified cod-end device (see Figure S1 for specifications).(b) Locations of sampling sites and stations between Waitematā Harbour and the Hauraki Gulf, New Zealand.
Figure S1 presents the IC device parts manufacturing dimensions, assembly, and cost specifications.To test and compare IC and CF methods, we used modified and conventional cod-ends attached to a Cruising Speed Net (CSN) that enables the collection of sub-surface marine seston material at up to 5 knots (von Ammon et al., 2020).

F
Doughnut diversity charts depicting the top 10 bacterial and eukaryotic family-level taxa between sample types (CF, conventional filtration; IC, in-water capture) and among genetic markers: (a) bacterial 16S rRNA, (b) 18S rRNA and (c) cytochrome oxidase I. *Unidentified bacterial or eukaryotic families.

F
I G U R E 4 Alpha and beta diversity measures of amplicon sequence variants (ASVs) across genetic markers, sample types and sites.(a) Box plot of observed richness between sampling types.(b-d) Non-metric multidimensional scaling (NMDS) plots displaying ASV community structure between sample types and among sites for (b) 16S, (c) 18S and (d) COI genes.